LG유플러스, MWC 2026 글로벌 AI 레드팀 챌린지 참가 - Korea IT Times
[AI] trusted ai solutions
|
|
🔬 연구
#ai 검증
#ai 레드팀
#gsma
#lg유플러스
#mwc 2026
#review
원문 출처: [AI] trusted ai solutions · Genesis Park에서 요약 및 분석
요약
LG Uplus가 내년 개최되는 'MWC 2026'에 참가하여 글로벌 AI 레드팀 챌린지에 도전한다는 소식입니다. 이 행사는 인공지능의 보안 취약점을 찾아내는 기술을 경쟁하는 자리로, 통신사의 선도적인 AI 안전 역량을 전 세계에 알리는 중요한 기회가 될 것으로 기대됩니다.
본문
LG Uplus announced it will participate in the Global AI Red Team Challenge, an AI validation program held during Mobile World Congress (MWC) 2026 in Barcelona, Spain. The event is organized by GSMA and is part of Talent Arena 2026, a developer and digital talent-focused program held alongside MWC. The challenge provides an environment for testing AI models developed by global telecom operators and technology companies under conditions that closely simulate real-world usage. LG Uplus will submit its in-house developed telecom-specialized generative AI model, ixi-GEN. Approximately 100 participants will act as “red team” evaluators, attempting to identify vulnerabilities in the AI system solely through prompt engineering. Unlike conventional hacking exercises, the challenge focuses on conversation-based attacks designed to bypass AI safety policies and safeguards. Participants will assess models across seven categories: jailbreak attempts, bias and discrimination, human rights violations, cyberattack facilitation, illegal content, misinformation, and response inconsistency. The evaluation criteria emphasize whether the AI provides information that should not be disclosed, rather than the factual accuracy of responses. Even if the content is factually incorrect, responses involving illegal activity or human rights violations will be treated as vulnerabilities. The approach aims to comprehensively assess potential risks associated with generative AI. The live challenge will run for 180 minutes, during which participants will be randomly assigned AI models and attempt to complete as many tasks as possible within the time limit. A panel of three judges will evaluate the results based on the number of successful attempts and overall success rates. LG Uplus said it plans to use the findings to objectively verify the safety of ixi-GEN and address identified vulnerabilities to further enhance the model. The company also intends to expand discussions on responsible AI usage and security through collaboration with global telecom operators and AI experts. Lee Hye-jin, head of Technology Strategy at LG Uplus, said that as AI adoption accelerates across telecom services and customer touchpoints, ensuring reliability and safety is as important as improving performance. She added that participation in the challenge will help strengthen the competitiveness of ixi-GEN and deliver trustworthy AI services to customers.
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유